Was this email forwarded to you? Sign up here for Telemetry.

Most robotics coverage lives at the level of the demonstration: a humanoid doing a backflip, a hand picking up something soft, an arm pulling one item off a shelf. The demo gets read as capability, and capability gets assumed to convert into deployment. Both inferences fail on contact with a real floor.

A humanoid that can backflip in a lab still cannot reach the inside of a car door on a moving line. The reason is not that the software needs another training run. The kinematic chain, the actuator bandwidth, and the controller were never built for that reach and that load, and changing any one of them is slow and costly.

The distance between a demo and a robot that does paid work for thousands of hours a year is a sequence of physical decisions: link lengths, actuator family, control loop, perception stack, integration with the equipment already on the floor. Every one of those decisions carries a penalty for which the demo doesn’t have to pay. (Note: revenue model and human interaction matter just as much; more below.)

I built Core Robotics to make these decisions legible, one substrate at a time. The figures are hand-drawn and every interactive is hand-tuned (I’m the guy in the header image, mid-sketch). The opening module of Core Robotics is completely free, and I’ve unlocked a few standout Interactives from the deeper, members-only modules, too. In there you’ll find Interactives I built to make the content ‘come alive’ on the page, as if they are micro-videogames. For example, Plot Your Robot on the Map: pick an archetype, from a warehouse autonomous mobile robot (AMR) to Mars rover, and discover which subsystem is its rate-limiter.

So as I put together each Interactive, I research the rhetorical “so what?” and strive to synthesize some takeaways for each one. But I don’t want folks to just take stuff at face value, so each Interactive has a special “Teardown” button. Click it and the Interactive flips inside-out to reveal the underlying assumptions and calculations that we’re doing “under the hood”!

Going back to the demos-are-cool-but-handwavy argument above, I pulled two from Core Robotics to highlight because they probe the illusion from opposite directions.

Map the Payload-Reach Envelope is an Interactive in which you can plot a job (how heavy, how far) against three real arms, including the common Universal Robots UR10e that carries 12.5 kg at 900mm of reach but only 8 kg once you extend it to its full 1,300mm, since mechanical ‘moment load’ eats capacity as the arm stretches out. If you push the sliders to 40 kg at 2,000mm you’ll see how no standard cobot covers the job! That blank region is where a mechanical redesign creates a market (and is not where marketing finds one).

The Choose Your Actuator Tradeoffs Interactive puts different actuator families (electric, hydraulic, pneumatic) side-by-side. You can see the physics that pushed Boston Dynamics to convert Atlas from hydraulic to electric in 2024, thereby trading peak force for a robot that does not bleed fluids on the floor. (To recreate the situation, open it and drag the force priority to its maximum and you’ll see hydraulics jump to the top of the ranking while the penalty radar balloons across weight, noise, maintenance, and a permanent leak liability!)

The two Interactives above interrogate the hardware. A third turns the same skepticism toward the company itself.

If you have an actual or hypothetical robotics startup in mind and have thought about what their product-service job demands, it’s a great time to probe who can actually build it. While most of the Core Robotics modules are reserved for our Operator and Principal members, I made public one more of these Interactives, a Company Scorecard, that exhibits how to structure an assessment: technology readiness, unit economics, integration maturity, team, defensibility, and timing. I pre-configured several companies, so, for example, you could load up 1X Technologies right after its NEO reveal; switch to Boston Dynamics and the axes shift. The NEO pitch promises a clean hexagon but diligence finds a dented polygon, with the space between being where most robotics venture losses live.

Each of our Interactives is built for our readers who wish to not only improve their intuition, but to formulate their own assumptions rather than solely accepting mine. So ‘teardown’ them, play with the dynamics, and watch the physics, business model, or secret sauce decide! (Many more Interactives are available as part of our membership program; check out our growing directory of Interactives.)

XPONENTIAL 2026 opened in Detroit earlier this month with Gov. Whitmer keynoting on the state's defense industrial base; the autonomy floor is now looking more like procurement-grade rather than demo-grade. Automate 2026 then lands in Chicago late-June, and the Hyundai Mobis × Boston Dynamics actuator framework announced at CES is the manipulation-stack story to watch translate into product on the floor. • DARPA CLARA awards its first cohort June 9, up to $2M per performer with mandated open-source release. Which formal-methods shops does the Pentagon now trust to certify autonomous systems? This is one of the gating answers beneath many well-we’d-buy-it-if-we-could-approve-it defense deals. • For lofting stuff up into space, the medium-lift launch cohort thinned in Q2: Neutron slipped to Q4 2026 after a tank burst in hydrostatic test, Eclipse moved to 2027, Terran R holds H2 2026, Nova still targets 2026 from SLC-14. • Since we’re on the Space domain topic, Vast slipped Haven-1 to NET Q1 2027, the second slip in eighteen months for the first standalone commercial station. The hull is welded and painted; alas, the gap between station press kits and on-orbit reality just widened by another year. This is useful context for the in-orbit-manipulation thesis our Core Robotics and Space for Earthlings pieces thread through.

No One Builds Alone.

/N1BA

Telemetry is written by JMill of The End Effector.